Filler words: what they actually do, and when they start to cost you

Every advice column says the same thing: get rid of «um». Almost nobody explains what «um» was doing there in the first place — and that is exactly the part that decides whether you can remove it without damaging the rest of your speech.

Fillers are not litter, they are signals

Herbert Clark and Jean Fox Tree argued that «uh» and «um» behave like words: speakers produce them deliberately enough that listeners can read them, and they carry a specific message — a delay is coming.

The two are not interchangeable in their account. The shorter one tends to announce a short delay; the longer one, a longer wait. A listener uses that the way a driver uses a turn signal: not as noise, but as information about what is about to happen.

This is also why every language has them, in its own shape. They are not a defect of English or of Russian or of Japanese. They are what human speech does when a mind is assembling a sentence in real time in front of another person.

What the listener does with them

A filler buys the speaker planning time without giving up the floor. That is the whole trade: you signal «still mine, still coming» and the other person waits instead of starting.

There is also evidence that a small disfluency can sharpen attention rather than blunt it — listeners appear to hold on to what follows a hesitation slightly better, presumably because the hesitation marks the coming word as effortful and therefore worth noticing.

So the honest version of the advice is not «fillers are bad». It is: fillers do a job, and above a certain rate they stop doing it.

When they start to cost you

The turn happens when the filler stops marking a real delay and becomes the seam between every clause. At that point it carries no information — it is just the sound your mouth makes while your head catches up.

Listeners notice density, not presence. One «um» in a paragraph is invisible. Six in a sentence is the only thing anyone will remember about the sentence.

The cost is largest exactly where people care most: presenting, interviewing, arguing a point. Not because the fillers make you unclear — usually they do not — but because they read as uncertainty about the content, and an audience that is deciding whether to believe you uses whatever it has.

Why «just stop» does not work

A filler is not a decision. It is a repair for a gap between the speed of your speech and the speed of your planning, and forbidding the repair does not close the gap.

That is why people who successfully suppress «um» often end up with something worse: a stretched vowel, a repeated word, a swallowed sentence, or a nervous run-on with no breathing at all. The pressure has to go somewhere.

The gap itself is the target. Slow the front of the sentence, and the filler has nothing to fill.

What to do instead

Replace, do not delete. The direct substitute for a filler is a silent pause — same length, same function, and it reads as composure instead of hesitation. Practising this feels absurd at first, because a pause always feels three times longer to the speaker than to the room.

Start sentences slower. Most fillers cluster in the first two or three words, where the sentence is least planned. Deliberately starting slow buys the planning time the filler was buying.

And measure before you judge. People are poor estimators of their own filler rate in both directions: some are convinced they say «um» constantly and do not, others have no idea. One recording, or a counter running while you talk, settles it in a day — and gives you a number to compare with next month instead of a feeling.